src/main/kotlin/com/project/movienight/adapters/metrics/BusinessMetricsService.kt

@@ -1,6 +1,7 @@
 package com.project.movienight.adapters.metrics
 
 import com.project.movienight.domain.model.JellyfinSyncSummary
+import com.project.movienight.domain.model.RecommendationEventType
 import io.micrometer.core.instrument.Counter
 import io.micrometer.core.instrument.MeterRegistry
 import io.micrometer.core.instrument.Timer
@@ -9,7 +10,7 @@ import java.util.concurrent.atomic.AtomicInteger
 
 @Service
 class BusinessMetricsService(
-    meterRegistry: MeterRegistry,
+    private val meterRegistry: MeterRegistry,
 ) {
     private val recommendationRequests: Counter = meterRegistry.counter("business_recommendation_requests_total")
     private val ratingsSubmitted: Counter = meterRegistry.counter("business_ratings_submitted_total")
@@ -36,6 +37,14 @@ class BusinessMetricsService(
         recommendationRequests.increment()
     }
 
+    fun recordRecommendationWeightsUpdated(eventType: RecommendationEventType) {
+        Counter
+            .builder("recommendation_weights_updated_total")
+            .tag("eventType", eventType.name)
+            .register(meterRegistry)
+            .increment()
+    }
+
     fun recordRatingSubmitted() {
         ratingsSubmitted.increment()
     }

src/main/kotlin/com/project/movienight/adapters/persistence/jdbc/RecommendationEventRepository.kt

@@ -0,0 +1,116 @@
+package com.project.movienight.adapters.persistence.jdbc
+
+import com.project.movienight.application.ports.output.RecommendationEventRepositoryPort
+import com.project.movienight.domain.model.RecommendationEvent
+import com.project.movienight.domain.model.RecommendationEventType
+import org.springframework.jdbc.core.JdbcTemplate
+import org.springframework.stereotype.Repository
+import java.sql.ResultSet
+import java.util.UUID
+
+@Repository
+class RecommendationEventRepository(
+    private val jdbc: JdbcTemplate,
+) : RecommendationEventRepositoryPort {
+    private val rowMapper = { rs: ResultSet, _: Int ->
+        RecommendationEvent(
+            id = UUID.fromString(rs.getString("id")),
+            userId = UUID.fromString(rs.getString("user_id")),
+            filmId = UUID.fromString(rs.getString("film_id")),
+            eventType = RecommendationEventType.valueOf(rs.getString("event_type")),
+            score = rs.getObject("score")?.let { (it as Number).toDouble() },
+            relevanceScore = rs.getObject("relevance_score")?.let { (it as Number).toDouble() },
+            qualityScore = rs.getObject("quality_score")?.let { (it as Number).toDouble() },
+            contextScore = rs.getObject("context_score")?.let { (it as Number).toDouble() },
+            noveltyScore = rs.getObject("novelty_score")?.let { (it as Number).toDouble() },
+            diversityScore = rs.getObject("diversity_score")?.let { (it as Number).toDouble() },
+            createdAt = rs.getTimestamp("created_at").toLocalDateTime(),
+        )
+    }
+
+    override fun save(event: RecommendationEvent): RecommendationEvent {
+        jdbc.update(
+            """
+            INSERT INTO recommendation_events (
+                id,
+                user_id,
+                film_id,
+                event_type,
+                score,
+                relevance_score,
+                quality_score,
+                context_score,
+                novelty_score,
+                diversity_score,
+                created_at
+            )
+            VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """.trimIndent(),
+            event.id,
+            event.userId,
+            event.filmId,
+            event.eventType.name,
+            event.score,
+            event.relevanceScore,
+            event.qualityScore,
+            event.contextScore,
+            event.noveltyScore,
+            event.diversityScore,
+            event.createdAt,
+        )
+        return event
+    }
+
+    override fun findByUserId(userId: UUID): List<RecommendationEvent> =
+        jdbc.query(
+            """
+            SELECT id,
+                   user_id,
+                   film_id,
+                   event_type,
+                   score,
+                   relevance_score,
+                   quality_score,
+                   context_score,
+                   novelty_score,
+                   diversity_score,
+                   created_at
+            FROM recommendation_events
+            WHERE user_id = ?
+            ORDER BY created_at DESC
+            """.trimIndent(),
+            rowMapper,
+            userId,
+        )
+
+    override fun findLatestRecommended(
+        userId: UUID,
+        filmId: UUID,
+    ): RecommendationEvent? =
+        jdbc
+            .query(
+                """
+                SELECT id,
+                       user_id,
+                       film_id,
+                       event_type,
+                       score,
+                       relevance_score,
+                       quality_score,
+                       context_score,
+                       novelty_score,
+                       diversity_score,
+                       created_at
+                FROM recommendation_events
+                WHERE user_id = ?
+                  AND film_id = ?
+                  AND event_type = ?
+                ORDER BY created_at DESC
+                LIMIT 1
+                """.trimIndent(),
+                rowMapper,
+                userId,
+                filmId,
+                RecommendationEventType.RECOMMENDED.name,
+            ).firstOrNull()
+}

src/main/kotlin/com/project/movienight/adapters/persistence/jdbc/UserRecommendationWeightsRepository.kt

@@ -0,0 +1,130 @@
+package com.project.movienight.adapters.persistence.jdbc
+
+import com.project.movienight.application.ports.output.UserRecommendationWeightsRepositoryPort
+import com.project.movienight.domain.model.UserRecommendationWeights
+import org.springframework.jdbc.core.JdbcTemplate
+import org.springframework.stereotype.Repository
+import java.sql.ResultSet
+import java.time.LocalDateTime
+import java.util.UUID
+
+@Repository
+class UserRecommendationWeightsRepository(
+    private val jdbc: JdbcTemplate,
+) : UserRecommendationWeightsRepositoryPort {
+    private val rowMapper = { rs: ResultSet, _: Int ->
+        UserRecommendationWeights(
+            userId = UUID.fromString(rs.getString("user_id")),
+            relevanceWeight = rs.getDouble("relevance_weight"),
+            qualityWeight = rs.getDouble("quality_weight"),
+            contextWeight = rs.getDouble("context_weight"),
+            noveltyWeight = rs.getDouble("novelty_weight"),
+            diversityWeight = rs.getDouble("diversity_weight"),
+            genreVectorWeight = rs.getDouble("genre_vector_weight"),
+            plotVectorWeight = rs.getDouble("plot_vector_weight"),
+            moodVectorWeight = rs.getDouble("mood_vector_weight"),
+            eraVectorWeight = rs.getDouble("era_vector_weight"),
+            peopleVectorWeight = rs.getDouble("people_vector_weight"),
+            contentTypeVectorWeight = rs.getDouble("content_type_vector_weight"),
+            updatedAt = rs.getTimestamp("updated_at").toLocalDateTime(),
+        )
+    }
+
+    override fun findByUserId(userId: UUID): UserRecommendationWeights? =
+        jdbc
+            .query(
+                """
+                SELECT user_id,
+                       relevance_weight,
+                       quality_weight,
+                       context_weight,
+                       novelty_weight,
+                       diversity_weight,
+                       genre_vector_weight,
+                       plot_vector_weight,
+                       mood_vector_weight,
+                       era_vector_weight,
+                       people_vector_weight,
+                       content_type_vector_weight,
+                       updated_at
+                FROM user_recommendation_weights
+                WHERE user_id = ?
+                """.trimIndent(),
+                rowMapper,
+                userId,
+            ).firstOrNull()
+
+    override fun save(weights: UserRecommendationWeights): UserRecommendationWeights {
+        val normalized = weights.normalized(updatedAt = LocalDateTime.now())
+        val updatedRows =
+            jdbc.update(
+                """
+                UPDATE user_recommendation_weights
+                SET relevance_weight = ?,
+                    quality_weight = ?,
+                    context_weight = ?,
+                    novelty_weight = ?,
+                    diversity_weight = ?,
+                    genre_vector_weight = ?,
+                    plot_vector_weight = ?,
+                    mood_vector_weight = ?,
+                    era_vector_weight = ?,
+                    people_vector_weight = ?,
+                    content_type_vector_weight = ?,
+                    updated_at = ?
+                WHERE user_id = ?
+                """.trimIndent(),
+                normalized.relevanceWeight,
+                normalized.qualityWeight,
+                normalized.contextWeight,
+                normalized.noveltyWeight,
+                normalized.diversityWeight,
+                normalized.genreVectorWeight,
+                normalized.plotVectorWeight,
+                normalized.moodVectorWeight,
+                normalized.eraVectorWeight,
+                normalized.peopleVectorWeight,
+                normalized.contentTypeVectorWeight,
+                normalized.updatedAt,
+                normalized.userId,
+            )
+
+        if (updatedRows == 0) {
+            jdbc.update(
+                """
+                INSERT INTO user_recommendation_weights (
+                    user_id,
+                    relevance_weight,
+                    quality_weight,
+                    context_weight,
+                    novelty_weight,
+                    diversity_weight,
+                    genre_vector_weight,
+                    plot_vector_weight,
+                    mood_vector_weight,
+                    era_vector_weight,
+                    people_vector_weight,
+                    content_type_vector_weight,
+                    updated_at
+                )
+                VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+                """.trimIndent(),
+                normalized.userId,
+                normalized.relevanceWeight,
+                normalized.qualityWeight,
+                normalized.contextWeight,
+                normalized.noveltyWeight,
+                normalized.diversityWeight,
+                normalized.genreVectorWeight,
+                normalized.plotVectorWeight,
+                normalized.moodVectorWeight,
+                normalized.eraVectorWeight,
+                normalized.peopleVectorWeight,
+                normalized.contentTypeVectorWeight,
+                normalized.updatedAt,
+            )
+        }
+
+        return normalized
+    }
+}

src/main/kotlin/com/project/movienight/adapters/web/RecommendationController.kt

@@ -1,34 +1,92 @@
 package com.project.movienight.adapters.web
 
+import com.project.movienight.adapters.web.dto.response.RecommendationEventResponse
+import com.project.movienight.adapters.web.dto.response.RecommendationResponse
+import com.project.movienight.application.ports.input.AcceptRecommendationCommand
+import com.project.movienight.application.ports.input.AcceptRecommendationUseCase
 import com.project.movienight.application.ports.input.GetRecommendationsUseCase
 import com.project.movienight.application.ports.input.RecommendationQuery
+import com.project.movienight.application.ports.input.RejectRecommendationCommand
+import com.project.movienight.application.ports.input.RejectRecommendationUseCase
+import com.project.movienight.config.JellyfinIntegrationProperties
 import com.project.movienight.domain.model.ContentType
-import com.project.movienight.domain.model.RecommendationResult
 import org.springframework.web.bind.annotation.GetMapping
 import org.springframework.web.bind.annotation.PathVariable
+import org.springframework.web.bind.annotation.PostMapping
 import org.springframework.web.bind.annotation.RequestMapping
 import org.springframework.web.bind.annotation.RequestParam
 import org.springframework.web.bind.annotation.RestController
+import java.net.URLEncoder
+import java.nio.charset.StandardCharsets
 import java.util.UUID
 
 @RestController
 @RequestMapping("/api/users/{userId}/recommendations")
 class RecommendationController(
     private val getRecommendationsUseCase: GetRecommendationsUseCase,
+    private val acceptRecommendationUseCase: AcceptRecommendationUseCase,
+    private val rejectRecommendationUseCase: RejectRecommendationUseCase,
+    private val jellyfinProperties: JellyfinIntegrationProperties,
 ) {
     @GetMapping
     fun recommend(
         @PathVariable userId: UUID,
         @RequestParam(required = false) contentType: String?,
         @RequestParam(required = false) mood: String?,
+        @RequestParam(required = false, defaultValue = "false") libraryOnly: Boolean,
         @RequestParam(required = false, defaultValue = "10") limit: Int,
-    ): List<RecommendationResult> =
-        getRecommendationsUseCase.recommend(
-            RecommendationQuery(
-                userId = userId,
-                contentType = contentType?.let { runCatching { ContentType.valueOf(it) }.getOrNull() },
-                mood = mood,
-                limit = limit,
+    ): List<RecommendationResponse> =
+        getRecommendationsUseCase
+            .recommend(
+                RecommendationQuery(
+                    userId = userId,
+                    contentType = contentType?.let { runCatching { ContentType.valueOf(it.uppercase()) }.getOrNull() },
+                    mood = mood,
+                    libraryOnly = libraryOnly,
+                    limit = limit,
+                ),
+            ).map { recommendation ->
+                RecommendationResponse.fromDomain(
+                    recommendation = recommendation,
+                    watchUrl = buildWatchUrl(recommendation.film.jellyfinItemId),
+                )
+            }
+
+    @PostMapping("/{filmId}/accept")
+    fun accept(
+        @PathVariable userId: UUID,
+        @PathVariable filmId: UUID,
+    ): RecommendationEventResponse =
+        RecommendationEventResponse.fromDomain(
+            acceptRecommendationUseCase.accept(
+                AcceptRecommendationCommand(
+                    userId = userId,
+                    filmId = filmId,
+                ),
             ),
         )
+
+    @PostMapping("/{filmId}/reject")
+    fun reject(
+        @PathVariable userId: UUID,
+        @PathVariable filmId: UUID,
+    ): RecommendationEventResponse =
+        RecommendationEventResponse.fromDomain(
+            rejectRecommendationUseCase.reject(
+                RejectRecommendationCommand(
+                    userId = userId,
+                    filmId = filmId,
+                ),
+            ),
+        )
+
+    private fun buildWatchUrl(jellyfinItemId: String?): String? {
+        if (jellyfinItemId.isNullOrBlank() || jellyfinProperties.webUrl.isBlank()) {
+            return null
+        }
+
+        val baseUrl = jellyfinProperties.webUrl.trimEnd('/')
+        val encodedItemId = URLEncoder.encode(jellyfinItemId, StandardCharsets.UTF_8)
+        return "$baseUrl/web/#/details?id=$encodedItemId"
+    }
 }

src/main/kotlin/com/project/movienight/adapters/web/RecommendationOnboardingController.kt

@@ -0,0 +1,52 @@
+package com.project.movienight.adapters.web
+
+import com.project.movienight.adapters.web.dto.request.RecommendationOnboardingRequest
+import com.project.movienight.adapters.web.dto.response.RecommendationOnboardingResponse
+import com.project.movienight.application.ports.input.CompleteRecommendationOnboardingCommand
+import com.project.movienight.application.ports.input.CompleteRecommendationOnboardingUseCase
+import com.project.movienight.domain.model.ContentType
+import com.project.movienight.domain.model.RecommendationStyle
+import org.springframework.web.bind.annotation.PathVariable
+import org.springframework.web.bind.annotation.PostMapping
+import org.springframework.web.bind.annotation.RequestBody
+import org.springframework.web.bind.annotation.RequestMapping
+import org.springframework.web.bind.annotation.RestController
+import java.util.Locale
+import java.util.UUID
+
+@RestController
+@RequestMapping("/api/users/{userId}/recommendation-onboarding")
+class RecommendationOnboardingController(
+    private val completeRecommendationOnboardingUseCase: CompleteRecommendationOnboardingUseCase,
+) {
+    @PostMapping
+    fun complete(
+        @PathVariable userId: UUID,
+        @RequestBody request: RecommendationOnboardingRequest,
+    ): RecommendationOnboardingResponse =
+        RecommendationOnboardingResponse.fromApplication(
+            completeRecommendationOnboardingUseCase.complete(
+                CompleteRecommendationOnboardingCommand(
+                    userId = userId,
+                    weightedGenres = request.weightedGenres,
+                    plotTypes = request.plotTypes,
+                    eras = request.eras,
+                    castAndDirectors = request.castAndDirectors,
+                    moods = request.moods,
+                    contentTypes = request.contentTypes.mapNotNull(::parseContentType),
+                    likedFilmIds = request.likedFilmIds,
+                    dislikedFilmIds = request.dislikedFilmIds,
+                    libraryFilmIds = request.libraryFilmIds,
+                    watchedFilmIds = request.watchedFilmIds,
+                    recommendationStyle = parseRecommendationStyle(request.recommendationStyle),
+                ),
+            ),
+        )
+
+    private fun parseContentType(value: String): ContentType? =
+        runCatching { ContentType.valueOf(value.uppercase(Locale.getDefault())) }.getOrNull()
+
+    private fun parseRecommendationStyle(value: String): RecommendationStyle =
+        runCatching { RecommendationStyle.valueOf(value.uppercase(Locale.getDefault())) }
+            .getOrDefault(RecommendationStyle.BALANCED)
+}

src/main/kotlin/com/project/movienight/adapters/web/UserRecommendationWeightsController.kt

@@ -0,0 +1,53 @@
+package com.project.movienight.adapters.web
+
+import com.project.movienight.adapters.web.dto.request.UpdateUserRecommendationWeightsRequest
+import com.project.movienight.adapters.web.dto.response.UserRecommendationWeightsResponse
+import com.project.movienight.application.ports.input.GetUserRecommendationWeightsUseCase
+import com.project.movienight.application.ports.input.UpdateUserRecommendationWeightsCommand
+import com.project.movienight.application.ports.input.UpdateUserRecommendationWeightsUseCase
+import org.springframework.web.bind.annotation.GetMapping
+import org.springframework.web.bind.annotation.PathVariable
+import org.springframework.web.bind.annotation.PutMapping
+import org.springframework.web.bind.annotation.RequestBody
+import org.springframework.web.bind.annotation.RequestMapping
+import org.springframework.web.bind.annotation.RestController
+import java.util.UUID
+
+@RestController
+@RequestMapping("/api/users/{userId}/recommendation-weights")
+class UserRecommendationWeightsController(
+    private val getUserRecommendationWeightsUseCase: GetUserRecommendationWeightsUseCase,
+    private val updateUserRecommendationWeightsUseCase: UpdateUserRecommendationWeightsUseCase,
+) {
+    @GetMapping
+    fun get(
+        @PathVariable userId: UUID,
+    ): UserRecommendationWeightsResponse =
+        UserRecommendationWeightsResponse.fromDomain(
+            getUserRecommendationWeightsUseCase.get(userId),
+        )
+
+    @PutMapping
+    fun update(
+        @PathVariable userId: UUID,
+        @RequestBody request: UpdateUserRecommendationWeightsRequest,
+    ): UserRecommendationWeightsResponse =
+        UserRecommendationWeightsResponse.fromDomain(
+            updateUserRecommendationWeightsUseCase.update(
+                UpdateUserRecommendationWeightsCommand(
+                    userId = userId,
+                    relevanceWeight = request.relevanceWeight,
+                    qualityWeight = request.qualityWeight,
+                    contextWeight = request.contextWeight,
+                    noveltyWeight = request.noveltyWeight,
+                    diversityWeight = request.diversityWeight,
+                    genreVectorWeight = request.genreVectorWeight,
+                    plotVectorWeight = request.plotVectorWeight,
+                    moodVectorWeight = request.moodVectorWeight,
+                    eraVectorWeight = request.eraVectorWeight,
+                    peopleVectorWeight = request.peopleVectorWeight,
+                    contentTypeVectorWeight = request.contentTypeVectorWeight,
+                ),
+            ),
+        )
+}

src/main/kotlin/com/project/movienight/adapters/web/dto/request/RecommendationOnboardingRequest.kt

@@ -0,0 +1,17 @@
+package com.project.movienight.adapters.web.dto.request
+
+import java.util.UUID
+
+data class RecommendationOnboardingRequest(
+    val weightedGenres: Map<String, Int> = emptyMap(),
+    val plotTypes: List<String> = emptyList(),
+    val eras: List<String> = emptyList(),
+    val castAndDirectors: List<String> = emptyList(),
+    val moods: List<String> = emptyList(),
+    val contentTypes: List<String> = emptyList(),
+    val likedFilmIds: List<UUID> = emptyList(),
+    val dislikedFilmIds: List<UUID> = emptyList(),
+    val libraryFilmIds: List<UUID> = emptyList(),
+    val watchedFilmIds: List<UUID> = emptyList(),
+    val recommendationStyle: String = "BALANCED",
+)

src/main/kotlin/com/project/movienight/adapters/web/dto/request/UpdateUserRecommendationWeightsRequest.kt

@@ -0,0 +1,15 @@
+package com.project.movienight.adapters.web.dto.request
+
+data class UpdateUserRecommendationWeightsRequest(
+    val relevanceWeight: Double,
+    val qualityWeight: Double,
+    val contextWeight: Double,
+    val noveltyWeight: Double,
+    val diversityWeight: Double,
+    val genreVectorWeight: Double,
+    val plotVectorWeight: Double,
+    val moodVectorWeight: Double,
+    val eraVectorWeight: Double,
+    val peopleVectorWeight: Double,
+    val contentTypeVectorWeight: Double,
+)

src/main/kotlin/com/project/movienight/adapters/web/dto/response/RecommendationEventResponse.kt

@@ -0,0 +1,37 @@
+package com.project.movienight.adapters.web.dto.response
+
+import com.project.movienight.domain.model.RecommendationEvent
+import com.project.movienight.domain.model.RecommendationEventType
+import java.time.LocalDateTime
+import java.util.UUID
+
+data class RecommendationEventResponse(
+    val id: UUID,
+    val userId: UUID,
+    val filmId: UUID,
+    val eventType: RecommendationEventType,
+    val score: Double?,
+    val relevanceScore: Double?,
+    val qualityScore: Double?,
+    val contextScore: Double?,
+    val noveltyScore: Double?,
+    val diversityScore: Double?,
+    val createdAt: LocalDateTime,
+) {
+    companion object {
+        fun fromDomain(event: RecommendationEvent): RecommendationEventResponse =
+            RecommendationEventResponse(
+                id = event.id,
+                userId = event.userId,
+                filmId = event.filmId,
+                eventType = event.eventType,
+                score = event.score,
+                relevanceScore = event.relevanceScore,
+                qualityScore = event.qualityScore,
+                contextScore = event.contextScore,
+                noveltyScore = event.noveltyScore,
+                diversityScore = event.diversityScore,
+                createdAt = event.createdAt,
+            )
+    }
+}

src/main/kotlin/com/project/movienight/adapters/web/dto/response/RecommendationOnboardingResponse.kt

@@ -0,0 +1,27 @@
+package com.project.movienight.adapters.web.dto.response
+
+import com.project.movienight.application.ports.input.RecommendationOnboardingResult
+import java.util.UUID
+
+data class RecommendationOnboardingResponse(
+    val userId: UUID,
+    val preferences: UserPreferencesResponse,
+    val weights: UserRecommendationWeightsResponse,
+    val likedFilmsCount: Int,
+    val dislikedFilmsCount: Int,
+    val libraryFilmsCount: Int,
+    val watchedFilmsCount: Int,
+) {
+    companion object {
+        fun fromApplication(result: RecommendationOnboardingResult): RecommendationOnboardingResponse =
+            RecommendationOnboardingResponse(
+                userId = result.userId,
+                preferences = UserPreferencesResponse.fromDomain(result.preferences),
+                weights = UserRecommendationWeightsResponse.fromDomain(result.weights),
+                likedFilmsCount = result.likedFilmsCount,
+                dislikedFilmsCount = result.dislikedFilmsCount,
+                libraryFilmsCount = result.libraryFilmsCount,
+                watchedFilmsCount = result.watchedFilmsCount,
+            )
+    }
+}

src/main/kotlin/com/project/movienight/adapters/web/dto/response/RecommendationResponse.kt

@@ -0,0 +1,32 @@
+package com.project.movienight.adapters.web.dto.response
+
+import com.project.movienight.domain.model.RecommendationResult
+import java.util.UUID
+
+data class RecommendationResponse(
+    val filmId: UUID,
+    val title: String,
+    val score: Double,
+    val reasons: List<String>,
+    val jellyfinItemId: String?,
+    val watchUrl: String?,
+    val film: FilmResponse,
+) {
+    companion object {
+        fun fromDomain(
+            recommendation: RecommendationResult,
+            watchUrl: String?,
+        ): RecommendationResponse {
+            val film = recommendation.film
+            return RecommendationResponse(
+                filmId = film.id,
+                title = film.title,
+                score = recommendation.score,
+                reasons = recommendation.reasons,
+                jellyfinItemId = film.jellyfinItemId,
+                watchUrl = watchUrl,
+                film = FilmResponse.fromDomain(film),
+            )
+        }
+    }
+}

src/main/kotlin/com/project/movienight/adapters/web/dto/response/UserRecommendationWeightsResponse.kt

@@ -0,0 +1,40 @@
+package com.project.movienight.adapters.web.dto.response
+
+import com.project.movienight.domain.model.UserRecommendationWeights
+import java.time.LocalDateTime
+import java.util.UUID
+
+data class UserRecommendationWeightsResponse(
+    val userId: UUID,
+    val relevanceWeight: Double,
+    val qualityWeight: Double,
+    val contextWeight: Double,
+    val noveltyWeight: Double,
+    val diversityWeight: Double,
+    val genreVectorWeight: Double,
+    val plotVectorWeight: Double,
+    val moodVectorWeight: Double,
+    val eraVectorWeight: Double,
+    val peopleVectorWeight: Double,
+    val contentTypeVectorWeight: Double,
+    val updatedAt: LocalDateTime,
+) {
+    companion object {
+        fun fromDomain(weights: UserRecommendationWeights): UserRecommendationWeightsResponse =
+            UserRecommendationWeightsResponse(
+                userId = weights.userId,
+                relevanceWeight = weights.relevanceWeight,
+                qualityWeight = weights.qualityWeight,
+                contextWeight = weights.contextWeight,
+                noveltyWeight = weights.noveltyWeight,
+                diversityWeight = weights.diversityWeight,
+                genreVectorWeight = weights.genreVectorWeight,
+                plotVectorWeight = weights.plotVectorWeight,
+                moodVectorWeight = weights.moodVectorWeight,
+                eraVectorWeight = weights.eraVectorWeight,
+                peopleVectorWeight = weights.peopleVectorWeight,
+                contentTypeVectorWeight = weights.contentTypeVectorWeight,
+                updatedAt = weights.updatedAt,
+            )
+    }
+}

src/main/kotlin/com/project/movienight/application/ports/input/GetRecommendationsUseCase.kt

@@ -1,6 +1,7 @@
 package com.project.movienight.application.ports.input
 
 import com.project.movienight.domain.model.ContentType
+import com.project.movienight.domain.model.RecommendationEvent
 import com.project.movienight.domain.model.RecommendationResult
 import java.util.UUID
 
@@ -12,5 +13,24 @@ data class RecommendationQuery(
     val userId: UUID,
     val contentType: ContentType? = null,
     val mood: String? = null,
+    val libraryOnly: Boolean = false,
     val limit: Int = 10,
 )
+
+interface AcceptRecommendationUseCase {
+    fun accept(command: AcceptRecommendationCommand): RecommendationEvent
+}
+
+data class AcceptRecommendationCommand(
+    val userId: UUID,
+    val filmId: UUID,
+)
+
+interface RejectRecommendationUseCase {
+    fun reject(command: RejectRecommendationCommand): RecommendationEvent
+}
+
+data class RejectRecommendationCommand(
+    val userId: UUID,
+    val filmId: UUID,
+)

src/main/kotlin/com/project/movienight/application/ports/input/RecommendationOnboardingUseCase.kt

@@ -0,0 +1,36 @@
+package com.project.movienight.application.ports.input
+
+import com.project.movienight.domain.model.ContentType
+import com.project.movienight.domain.model.RecommendationStyle
+import com.project.movienight.domain.model.UserPreferences
+import com.project.movienight.domain.model.UserRecommendationWeights
+import java.util.UUID
+
+interface CompleteRecommendationOnboardingUseCase {
+    fun complete(command: CompleteRecommendationOnboardingCommand): RecommendationOnboardingResult
+}
+
+data class CompleteRecommendationOnboardingCommand(
+    val userId: UUID,
+    val weightedGenres: Map<String, Int> = emptyMap(),
+    val plotTypes: List<String> = emptyList(),
+    val eras: List<String> = emptyList(),
+    val castAndDirectors: List<String> = emptyList(),
+    val moods: List<String> = emptyList(),
+    val contentTypes: List<ContentType> = emptyList(),
+    val likedFilmIds: List<UUID> = emptyList(),
+    val dislikedFilmIds: List<UUID> = emptyList(),
+    val libraryFilmIds: List<UUID> = emptyList(),
+    val watchedFilmIds: List<UUID> = emptyList(),
+    val recommendationStyle: RecommendationStyle = RecommendationStyle.BALANCED,
+)
+
+data class RecommendationOnboardingResult(
+    val userId: UUID,
+    val preferences: UserPreferences,
+    val weights: UserRecommendationWeights,
+    val likedFilmsCount: Int,
+    val dislikedFilmsCount: Int,
+    val libraryFilmsCount: Int,
+    val watchedFilmsCount: Int,
+)

src/main/kotlin/com/project/movienight/application/ports/input/UserRecommendationWeightsUseCase.kt

@@ -0,0 +1,27 @@
+package com.project.movienight.application.ports.input
+
+import com.project.movienight.domain.model.UserRecommendationWeights
+import java.util.UUID
+
+interface GetUserRecommendationWeightsUseCase {
+    fun get(userId: UUID): UserRecommendationWeights
+}
+
+interface UpdateUserRecommendationWeightsUseCase {
+    fun update(command: UpdateUserRecommendationWeightsCommand): UserRecommendationWeights
+}
+
+data class UpdateUserRecommendationWeightsCommand(
+    val userId: UUID,
+    val relevanceWeight: Double,
+    val qualityWeight: Double,
+    val contextWeight: Double,
+    val noveltyWeight: Double,
+    val diversityWeight: Double,
+    val genreVectorWeight: Double,
+    val plotVectorWeight: Double,
+    val moodVectorWeight: Double,
+    val eraVectorWeight: Double,
+    val peopleVectorWeight: Double,
+    val contentTypeVectorWeight: Double,
+)

src/main/kotlin/com/project/movienight/application/ports/output/RecommendationEventRepositoryPort.kt

@@ -0,0 +1,15 @@
+package com.project.movienight.application.ports.output
+
+import com.project.movienight.domain.model.RecommendationEvent
+import java.util.UUID
+
+interface RecommendationEventRepositoryPort {
+    fun save(event: RecommendationEvent): RecommendationEvent
+
+    fun findByUserId(userId: UUID): List<RecommendationEvent>
+
+    fun findLatestRecommended(
+        userId: UUID,
+        filmId: UUID,
+    ): RecommendationEvent?
+}

src/main/kotlin/com/project/movienight/application/ports/output/UserRecommendationWeightsRepositoryPort.kt

@@ -0,0 +1,10 @@
+package com.project.movienight.application.ports.output
+
+import com.project.movienight.domain.model.UserRecommendationWeights
+import java.util.UUID
+
+interface UserRecommendationWeightsRepositoryPort {
+    fun findByUserId(userId: UUID): UserRecommendationWeights?
+
+    fun save(weights: UserRecommendationWeights): UserRecommendationWeights
+}

src/main/kotlin/com/project/movienight/application/services/RecommendationOnboardingService.kt

@@ -0,0 +1,150 @@
+package com.project.movienight.application.services
+
+import com.project.movienight.application.ports.input.CompleteRecommendationOnboardingCommand
+import com.project.movienight.application.ports.input.CompleteRecommendationOnboardingUseCase
+import com.project.movienight.application.ports.input.RecommendationOnboardingResult
+import com.project.movienight.application.ports.output.FilmLibraryRepositoryPort
+import com.project.movienight.application.ports.output.FilmRatingRepositoryPort
+import com.project.movienight.application.ports.output.FilmRepositoryPort
+import com.project.movienight.application.ports.output.IdGenerator
+import com.project.movienight.application.ports.output.UserPreferencesRepositoryPort
+import com.project.movienight.application.ports.output.UserRecommendationWeightsRepositoryPort
+import com.project.movienight.application.ports.output.UserRepositoryPort
+import com.project.movienight.domain.exception.EntityNotFoundException
+import com.project.movienight.domain.model.FilmLibrary
+import com.project.movienight.domain.model.FilmRating
+import com.project.movienight.domain.model.UserPreferences
+import com.project.movienight.domain.model.UserRecommendationWeights
+import org.springframework.stereotype.Service
+import java.time.LocalDateTime
+import java.util.UUID
+
+@Service
+class RecommendationOnboardingService(
+    private val userRepository: UserRepositoryPort,
+    private val filmRepository: FilmRepositoryPort,
+    private val userPreferencesRepository: UserPreferencesRepositoryPort,
+    private val filmRatingRepository: FilmRatingRepositoryPort,
+    private val filmLibraryRepository: FilmLibraryRepositoryPort,
+    private val userRecommendationWeightsRepository: UserRecommendationWeightsRepositoryPort,
+    private val idGenerator: IdGenerator,
+) : CompleteRecommendationOnboardingUseCase {
+    override fun complete(command: CompleteRecommendationOnboardingCommand): RecommendationOnboardingResult {
+        userRepository.findById(command.userId)
+            ?: throw EntityNotFoundException(entity = "User", id = command.userId.toString())
+
+        val filmIds =
+            (
+                command.likedFilmIds +
+                    command.dislikedFilmIds +
+                    command.libraryFilmIds +
+                    command.watchedFilmIds
+            ).distinct()
+        ensureFilmsExist(filmIds)
+
+        val preferences =
+            userPreferencesRepository.save(
+                UserPreferences(
+                    userId = command.userId,
+                    weightedGenres = command.weightedGenres,
+                    plotTypes = command.plotTypes,
+                    eras = command.eras,
+                    castAndDirectors = command.castAndDirectors,
+                    moods = command.moods,
+                    contentTypes = command.contentTypes,
+                ),
+            )
+
+        command.likedFilmIds.distinct().forEach { filmId ->
+            saveRating(userId = command.userId, filmId = filmId, score = LIKED_SCORE, note = ONBOARDING_LIKED_NOTE)
+        }
+        command.dislikedFilmIds.distinct().forEach { filmId ->
+            saveRating(
+                userId = command.userId,
+                filmId = filmId,
+                score = DISLIKED_SCORE,
+                note = ONBOARDING_DISLIKED_NOTE,
+            )
+        }
+        command.libraryFilmIds.distinct().forEach { filmId ->
+            saveLibraryEntry(userId = command.userId, filmId = filmId, isViewed = false)
+        }
+        command.watchedFilmIds.distinct().forEach { filmId ->
+            saveLibraryEntry(userId = command.userId, filmId = filmId, isViewed = true)
+        }
+
+        val weights =
+            userRecommendationWeightsRepository.save(
+                UserRecommendationWeights.forStyle(
+                    userId = command.userId,
+                    style = command.recommendationStyle,
+                ),
+            )
+
+        return RecommendationOnboardingResult(
+            userId = command.userId,
+            preferences = preferences,
+            weights = weights,
+            likedFilmsCount = command.likedFilmIds.distinct().size,
+            dislikedFilmsCount = command.dislikedFilmIds.distinct().size,
+            libraryFilmsCount = command.libraryFilmIds.distinct().size,
+            watchedFilmsCount = command.watchedFilmIds.distinct().size,
+        )
+    }
+
+    private fun ensureFilmsExist(filmIds: List<UUID>) {
+        filmIds.forEach { filmId ->
+            filmRepository.findById(filmId)
+                ?: throw EntityNotFoundException(entity = "Film", id = filmId.toString())
+        }
+    }
+
+    private fun saveRating(
+        userId: UUID,
+        filmId: UUID,
+        score: Int,
+        note: String,
+    ): FilmRating {
+        val now = LocalDateTime.now()
+        val existing = filmRatingRepository.findByUserIdAndFilmId(userId, filmId)
+        return filmRatingRepository.save(
+            existing?.copy(score = score, note = note, updatedAt = now)
+                ?: FilmRating(
+                    id = idGenerator.generateId(),
+                    userId = userId,
+                    filmId = filmId,
+                    score = score,
+                    note = note,
+                    createdAt = now,
+                    updatedAt = now,
+                ),
+        )
+    }
+
+    private fun saveLibraryEntry(
+        userId: UUID,
+        filmId: UUID,
+        isViewed: Boolean,
+    ): FilmLibrary {
+        val watchedAt = LocalDateTime.now().takeIf { isViewed }
+        val existing = filmLibraryRepository.findByUserIdAndFilmId(userId, filmId)
+        return filmLibraryRepository.save(
+            existing?.copy(isViewed = isViewed, watchedAt = watchedAt)
+                ?: FilmLibrary(
+                    id = idGenerator.generateId(),
+                    userId = userId,
+                    filmId = filmId,
+                    comment = null,
+                    isViewed = isViewed,
+                    watchedAt = watchedAt,
+                ),
+        )
+    }
+
+    private companion object {
+        private const val LIKED_SCORE = 10
+        private const val DISLIKED_SCORE = 2
+        private const val ONBOARDING_LIKED_NOTE = "Onboarding liked"
+        private const val ONBOARDING_DISLIKED_NOTE = "Onboarding disliked"
+    }
+}

src/main/kotlin/com/project/movienight/application/services/RecommendationService.kt

@@ -1,16 +1,35 @@
 package com.project.movienight.application.services
 
 import com.project.movienight.adapters.metrics.BusinessMetricsService
+import com.project.movienight.application.ports.input.AcceptRecommendationCommand
+import com.project.movienight.application.ports.input.AcceptRecommendationUseCase
 import com.project.movienight.application.ports.input.GetRecommendationsUseCase
 import com.project.movienight.application.ports.input.RecommendationQuery
+import com.project.movienight.application.ports.input.RejectRecommendationCommand
+import com.project.movienight.application.ports.input.RejectRecommendationUseCase
 import com.project.movienight.application.ports.output.FilmLibraryRepositoryPort
 import com.project.movienight.application.ports.output.FilmRatingRepositoryPort
 import com.project.movienight.application.ports.output.FilmRepositoryPort
+import com.project.movienight.application.ports.output.IdGenerator
+import com.project.movienight.application.ports.output.RecommendationEventRepositoryPort
 import com.project.movienight.application.ports.output.UserPreferencesRepositoryPort
-import com.project.movienight.domain.model.ContentType
+import com.project.movienight.application.ports.output.UserRecommendationWeightsRepositoryPort
+import com.project.movienight.application.ports.output.UserRepositoryPort
+import com.project.movienight.domain.exception.EntityNotFoundException
 import com.project.movienight.domain.model.Film
+import com.project.movienight.domain.model.FilmLibrary
+import com.project.movienight.domain.model.FilmRating
+import com.project.movienight.domain.model.RecommendationEvent
+import com.project.movienight.domain.model.RecommendationEventType
 import com.project.movienight.domain.model.RecommendationResult
+import com.project.movienight.domain.model.UserPreferences
+import com.project.movienight.domain.model.UserRecommendationWeights
+import org.slf4j.LoggerFactory
 import org.springframework.stereotype.Service
+import java.time.LocalDateTime
+import java.util.Locale
+import java.util.UUID
+import kotlin.math.sqrt
 
 @Service
 class RecommendationService(
@@ -18,100 +37,638 @@ class RecommendationService(
     private val filmLibraryRepository: FilmLibraryRepositoryPort,
     private val filmRatingRepository: FilmRatingRepositoryPort,
     private val userPreferencesRepository: UserPreferencesRepositoryPort,
+    private val userRepository: UserRepositoryPort,
+    private val recommendationEventRepository: RecommendationEventRepositoryPort,
+    private val userRecommendationWeightsRepository: UserRecommendationWeightsRepositoryPort,
+    private val idGenerator: IdGenerator,
     private val businessMetricsService: BusinessMetricsService,
-) : GetRecommendationsUseCase {
+) : GetRecommendationsUseCase,
+    AcceptRecommendationUseCase,
+    RejectRecommendationUseCase {
+    private val log = LoggerFactory.getLogger(javaClass)
+
     override fun recommend(query: RecommendationQuery): List<RecommendationResult> {
         businessMetricsService.recordRecommendationRequest()
-        val preferences = userPreferencesRepository.findByUserId(query.userId)
-        val ratings = filmRatingRepository.findByUserId(query.userId).associateBy { it.filmId }
-        val watchedFilmIds =
-            filmLibraryRepository
-                .findAll()
-                .filter {
-                    it.userId == query.userId && it.isViewed
-                }.map { it.filmId }
-                .toSet()
+        userRepository.findById(query.userId)
+            ?: throw EntityNotFoundException(entity = "User", id = query.userId.toString())
 
-        return filmRepository
-            .findAll()
-            .asSequence()
-            .filter { film -> query.contentType == null || film.contentType == query.contentType }
-            .map { film ->
-                scoreFilm(film, query.mood, preferences, ratings[film.id] != null, watchedFilmIds.contains(film.id))
-            }.sortedByDescending { it.score }
-            .take(query.limit.coerceAtLeast(1))
-            .toList()
+        val preferences = userPreferencesRepository.findByUserId(query.userId)
+        val ratings = filmRatingRepository.findByUserId(query.userId)
+        val libraryEntries = filmLibraryRepository.findAll().filter { it.userId == query.userId }
+        val libraryFilmIds = libraryEntries.map { it.filmId }.toSet()
+        val watchedFilmIds = libraryEntries.filter { it.isViewed }.map { it.filmId }.toSet()
+        val films = filmRepository.findAll()
+        val filmsById = films.associateBy { it.id }
+        val weights = findWeights(query.userId)
+        val userProfile = buildUserProfile(preferences, ratings, libraryEntries, filmsById, weights)
+
+        val candidates =
+            films
+                .asSequence()
+                .filter { film -> query.contentType == null || film.contentType == query.contentType }
+                .filter { film -> film.id !in watchedFilmIds }
+                .filter { film -> !query.libraryOnly || film.id in libraryFilmIds }
+                .toList()
+        val scoredCandidates =
+            candidates.map { film ->
+                scoreFilm(film, query, preferences, userProfile, film.id in libraryFilmIds, weights)
+            }
+        val recommendationComparator =
+            compareByDescending<ScoredRecommendation> { it.result.score }.thenBy {
+                it.result.film.title
+            }
+        val scoredRecommendations =
+            scoredCandidates
+                .sortedWith(recommendationComparator)
+                .take(query.limit.coerceAtLeast(1))
+
+        scoredRecommendations.forEach { recommendation ->
+            saveEvent(
+                userId = query.userId,
+                filmId = recommendation.result.film.id,
+                eventType = RecommendationEventType.RECOMMENDED,
+                score = recommendation.result.score,
+                relevanceScore = recommendation.relevanceScore,
+                qualityScore = recommendation.qualityScore,
+                contextScore = recommendation.contextScore,
+                noveltyScore = recommendation.noveltyScore,
+                diversityScore = recommendation.diversityScore,
+            )
+        }
+
+        log.info(
+            RECOMMENDATION_COMPLETED_LOG,
+            query.userId,
+            query.contentType,
+            !query.mood.isNullOrBlank(),
+            query.libraryOnly,
+            query.limit,
+            candidates.size,
+            scoredRecommendations.size,
+        )
+        if (log.isDebugEnabled) {
+            log.debug(
+                "Recommendation top results: userId='{}', results='{}'",
+                query.userId,
+                scoredRecommendations.joinToString(separator = ",") { "${it.result.film.id}:${it.result.score}" },
+            )
+        }
+
+        return scoredRecommendations.map { it.result }
+    }
+
+    override fun accept(command: AcceptRecommendationCommand): RecommendationEvent =
+        saveFeedbackEvent(
+            userId = command.userId,
+            filmId = command.filmId,
+            eventType = RecommendationEventType.ACCEPTED,
+        )
+
+    override fun reject(command: RejectRecommendationCommand): RecommendationEvent =
+        saveFeedbackEvent(
+            userId = command.userId,
+            filmId = command.filmId,
+            eventType = RecommendationEventType.REJECTED,
+        )
+
+    private fun saveFeedbackEvent(
+        userId: UUID,
+        filmId: UUID,
+        eventType: RecommendationEventType,
+    ): RecommendationEvent {
+        userRepository.findById(userId)
+            ?: throw EntityNotFoundException(entity = "User", id = userId.toString())
+        filmRepository.findById(filmId)
+            ?: throw EntityNotFoundException(entity = "Film", id = filmId.toString())
+
+        val lastRecommendation = recommendationEventRepository.findLatestRecommended(userId, filmId)
+        val event =
+            saveEvent(
+                userId = userId,
+                filmId = filmId,
+                eventType = eventType,
+                score = lastRecommendation?.score,
+                relevanceScore = lastRecommendation?.relevanceScore,
+                qualityScore = lastRecommendation?.qualityScore,
+                contextScore = lastRecommendation?.contextScore,
+                noveltyScore = lastRecommendation?.noveltyScore,
+                diversityScore = lastRecommendation?.diversityScore,
+            )
+
+        if (lastRecommendation != null) {
+            updateRecommendationWeights(
+                userId = userId,
+                eventType = eventType,
+                recommendation = lastRecommendation,
+            )
+        } else {
+            log.info(
+                "Recommendation feedback saved without weight update: userId='{}', filmId='{}', eventType='{}'",
+                userId,
+                filmId,
+                eventType,
+            )
+        }
+
+        log.info(
+            RECOMMENDATION_FEEDBACK_SAVED_LOG,
+            userId,
+            filmId,
+            eventType,
+        )
+
+        return event
+    }
+
+    private fun saveEvent(
+        userId: UUID,
+        filmId: UUID,
+        eventType: RecommendationEventType,
+        score: Double?,
+        relevanceScore: Double? = null,
+        qualityScore: Double? = null,
+        contextScore: Double? = null,
+        noveltyScore: Double? = null,
+        diversityScore: Double? = null,
+    ): RecommendationEvent =
+        recommendationEventRepository.save(
+            RecommendationEvent(
+                id = idGenerator.generateId(),
+                userId = userId,
+                filmId = filmId,
+                eventType = eventType,
+                score = score,
+                relevanceScore = relevanceScore,
+                qualityScore = qualityScore,
+                contextScore = contextScore,
+                noveltyScore = noveltyScore,
+                diversityScore = diversityScore,
+                createdAt = LocalDateTime.now(),
+            ),
+        )
+
+    private fun findWeights(userId: UUID): UserRecommendationWeights =
+        (
+            userRecommendationWeightsRepository.findByUserId(userId)
+                ?: UserRecommendationWeights.defaultFor(userId)
+        ).normalized()
+
+    private fun updateRecommendationWeights(
+        userId: UUID,
+        eventType: RecommendationEventType,
+        recommendation: RecommendationEvent,
+    ) {
+        val current = findWeights(userId)
+        val contributions = scoreContributions(recommendation, current) ?: return
+        val direction =
+            when (eventType) {
+                RecommendationEventType.ACCEPTED -> 1.0
+                RecommendationEventType.REJECTED -> -1.0
+                RecommendationEventType.RECOMMENDED -> return
+            }
+
+        val updated =
+            current
+                .copy(
+                    relevanceWeight = current.relevanceWeight + direction * LEARNING_RATE * contributions.relevance,
+                    qualityWeight = current.qualityWeight + direction * LEARNING_RATE * contributions.quality,
+                    contextWeight = current.contextWeight + direction * LEARNING_RATE * contributions.context,
+                    noveltyWeight = current.noveltyWeight + direction * LEARNING_RATE * contributions.novelty,
+                    diversityWeight = current.diversityWeight + direction * LEARNING_RATE * contributions.diversity,
+                ).normalized(updatedAt = LocalDateTime.now())
+
+        val saved = userRecommendationWeightsRepository.save(updated)
+        businessMetricsService.recordRecommendationWeightsUpdated(eventType)
+        log.info(
+            RECOMMENDATION_WEIGHTS_UPDATED_LOG,
+            userId,
+            eventType,
+            current.hashCode(),
+            saved.hashCode(),
+        )
+    }
+
+    private fun scoreContributions(
+        recommendation: RecommendationEvent,
+        weights: UserRecommendationWeights,
+    ): ScoreContributions? {
+        val rawContributions =
+            listOf(
+                weights.relevanceWeight to recommendation.relevanceScore,
+                weights.qualityWeight to recommendation.qualityScore,
+                weights.contextWeight to recommendation.contextScore,
+                weights.noveltyWeight to recommendation.noveltyScore,
+                weights.diversityWeight to recommendation.diversityScore,
+            ).map { (weight, score) ->
+                weight * (score?.takeIf { value -> value.isFinite() }?.coerceAtLeast(0.0) ?: 0.0)
+            }
+        val total = rawContributions.sum()
+        if (total <= 0.0) {
+            return null
+        }
+        return ScoreContributions(
+            relevance = rawContributions[0] / total,
+            quality = rawContributions[1] / total,
+            context = rawContributions[2] / total,
+            novelty = rawContributions[3] / total,
+            diversity = rawContributions[4] / total,
+        )
+    }
+
+    private fun buildUserProfile(
+        preferences: UserPreferences?,
+        ratings: List<FilmRating>,
+        libraryEntries: List<FilmLibrary>,
+        filmsById: Map<UUID, Film>,
+        weights: UserRecommendationWeights,
+    ): SparseVector {
+        val profile = MutableSparseVector()
+
+        preferences?.weightedGenres.orEmpty().forEach { (genre, weight) ->
+            profile.add(feature("genre", genre), weight.coerceAtLeast(1).toDouble() / MAX_PREFERENCE_WEIGHT)
+        }
+        preferences?.plotTypes.orEmpty().forEach { plotType ->
+            tokenize(plotType).forEach { profile.add(feature("plot", it), PREFERENCE_PLOT_WEIGHT) }
+        }
+        preferences?.eras.orEmpty().forEach { profile.add(feature("era", it), PREFERENCE_ERA_WEIGHT) }
+        preferences?.castAndDirectors.orEmpty().forEach { profile.add(feature("person", it), PREFERENCE_PERSON_WEIGHT) }
+        preferences?.moods.orEmpty().forEach { profile.add(feature("mood", it), PREFERENCE_MOOD_WEIGHT) }
+        preferences
+            ?.contentTypes
+            .orEmpty()
+            .forEach {
+                profile.add(
+                    feature("type", it.name),
+                    PREFERENCE_CONTENT_TYPE_WEIGHT,
+                )
+            }
+
+        ratings.forEach { rating ->
+            val film = filmsById[rating.filmId] ?: return@forEach
+            val signal = ratingSignal(rating.score)
+            profile.add(buildFilmVector(film, weights).scale(signal))
+        }
+
+        libraryEntries.filterNot { it.isViewed }.forEach { entry ->
+            val film = filmsById[entry.filmId] ?: return@forEach
+            profile.add(buildFilmVector(film, weights).scale(LIBRARY_SIGNAL_WEIGHT))
+        }
+
+        return profile.toSparseVector()
     }
 
     private fun scoreFilm(
         film: Film,
-        mood: String?,
-        preferences: com.project.movienight.domain.model.UserPreferences?,
-        hasUserRating: Boolean,
-        watched: Boolean,
-    ): RecommendationResult {
-        var score = 0.0
+        query: RecommendationQuery,
+        preferences: UserPreferences?,
+        userProfile: SparseVector,
+        inLibrary: Boolean,
+        weights: UserRecommendationWeights,
+    ): ScoredRecommendation {
         val reasons = mutableListOf<String>()
+        val filmVector = buildFilmVector(film, weights)
+        val preferenceScore = cosineSimilarity(userProfile, filmVector)
+        val qualityScore = qualityScore(film)
+        val contextScore = contextScore(film, query, preferences)
+        val noveltyScore = if (inLibrary) LIBRARY_NOVELTY_SCORE else CATALOG_NOVELTY_SCORE
+        val diversityScore = diversityScore(film, preferences)
+        val score =
+            weights.relevanceWeight * preferenceScore +
+                weights.qualityWeight * qualityScore +
+                weights.contextWeight * contextScore +
+                weights.noveltyWeight * noveltyScore +
+                weights.diversityWeight * diversityScore
 
-        preferences?.contentTypes?.let {
-            if (it.isEmpty() || it.contains(film.contentType)) {
-                score += 2.0
-                reasons += "Matches content preference"
+        if (preferenceScore > STRONG_REASON_THRESHOLD) {
+            reasons += "Similar to user preferences and rating history"
+        }
+        matchingGenres(film, preferences).take(MAX_REASON_ITEMS).forEach { genre ->
+            reasons += "Matches preferred genre: $genre"
+        }
+        matchingPeople(film, preferences).take(MAX_REASON_ITEMS).forEach { person ->
+            reasons += "Matches preferred cast or director: $person"
+        }
+        query.mood?.takeIf { inferredMoods(film).contains(normalize(it)) }?.let { mood ->
+            reasons += "Matches requested mood: $mood"
+        }
+        film.releaseYear?.let { year ->
+            if (preferences?.eras.orEmpty().any { normalize(it) == normalize(decadeOf(year)) }) {
+                reasons += "Matches preferred era: ${decadeOf(year)}"
             }
         }
-
-        preferences?.weightedGenres?.forEach { (genre, weight) ->
-            if (film.genres.any { it.equals(genre, ignoreCase = true) }) {
-                score += weight
-                reasons += "Matches genre $genre"
-            }
+        if (qualityScore >= QUALITY_REASON_THRESHOLD) {
+            reasons += "High rating signal"
         }
-
-        preferences?.castAndDirectors?.forEach { favorite ->
-            val found =
-                film.cast.any { it.equals(favorite, ignoreCase = true) } ||
-                    film.directors.any { it.equals(favorite, ignoreCase = true) }
-            if (found) {
-                score += 1.5
-                reasons += "Matches favorite creator or cast member $favorite"
-            }
-        }
-
-        preferences?.moods?.forEach { preferredMood ->
-            if (mood != null && preferredMood.equals(mood, ignoreCase = true)) {
-                score += 1.25
-                reasons += "Matches requested mood $mood"
-            }
-        }
-
-        film.imdbRating?.let {
-            score += it / 2.0
-            reasons += "Strong IMDb signal"
-        }
-
-        film.platformRating?.let {
-            score += it
-            reasons += "Strong platform signal"
-        }
-
-        if (hasUserRating) {
-            score += 2.0
-            reasons += "User has already rated similar content"
-        }
-
-        if (watched) {
-            score -= 3.0
-            reasons += "Already watched"
-        }
-
-        if (mood != null && film.title.contains(mood, ignoreCase = true)) {
-            score += 0.5
+        if (inLibrary) {
+            reasons += "Already in user library"
         }
 
         if (reasons.isEmpty()) {
-            reasons += "Baseline recommendation from library catalog"
+            reasons += "Baseline recommendation from catalog quality"
         }
 
-        return RecommendationResult(film = film, score = score, reasons = reasons)
+        return ScoredRecommendation(
+            result = RecommendationResult(film = film, score = roundScore(score), reasons = reasons.distinct()),
+            relevanceScore = preferenceScore,
+            qualityScore = qualityScore,
+            contextScore = contextScore,
+            noveltyScore = noveltyScore,
+            diversityScore = diversityScore,
+        )
+    }
+
+    private fun buildFilmVector(
+        film: Film,
+        weights: UserRecommendationWeights,
+    ): SparseVector {
+        val vector = MutableSparseVector()
+        val normalizedGenres = film.genres.map(::normalize).filter { it.isNotBlank() }
+        val plotTokens = tokenize("${film.title} ${film.description}")
+        val moods = inferredMoods(film)
+        val people = (film.directors + film.cast).map(::normalize).filter { it.isNotBlank() }
+
+        vector.add(feature("type", film.contentType.name), weights.contentTypeVectorWeight)
+        distribute(vector, "genre", normalizedGenres, weights.genreVectorWeight)
+        distribute(vector, "plot", plotTokens, weights.plotVectorWeight)
+        distribute(vector, "mood", moods, weights.moodVectorWeight)
+        film.releaseYear?.let { vector.add(feature("era", decadeOf(it)), weights.eraVectorWeight) }
+        distribute(vector, "person", people, weights.peopleVectorWeight)
+
+        return vector.toSparseVector()
+    }
+
+    private fun contextScore(
+        film: Film,
+        query: RecommendationQuery,
+        preferences: UserPreferences?,
+    ): Double {
+        var score = 0.0
+        var checks = 0
+
+        query.mood?.let {
+            checks += 1
+            if (inferredMoods(film).contains(normalize(it))) {
+                score += 1.0
+            }
+        }
+        preferences?.contentTypes?.takeIf { it.isNotEmpty() }?.let {
+            checks += 1
+            if (film.contentType in it) {
+                score += 1.0
+            }
+        }
+        preferences?.eras?.takeIf { it.isNotEmpty() }?.let { eras ->
+            film.releaseYear?.let {
+                checks += 1
+                if (eras.any { era -> normalize(era) == normalize(decadeOf(it)) }) {
+                    score += 1.0
+                }
+            }
+        }
+
+        return if (checks == 0) BASE_CONTEXT_SCORE else score / checks
+    }
+
+    private fun qualityScore(film: Film): Double {
+        val normalizedRatings =
+            listOfNotNull(
+                film.imdbRating?.let { normalizeRating(it) },
+                film.platformRating?.let { normalizeRating(it) },
+            )
+        return normalizedRatings.averageOrNull() ?: BASE_QUALITY_SCORE
+    }
+
+    private fun diversityScore(
+        film: Film,
+        preferences: UserPreferences?,
+    ): Double {
+        val preferredGenres =
+            preferences
+                ?.weightedGenres
+                .orEmpty()
+                .keys
+                .map(::normalize)
+                .toSet()
+        val filmGenres = film.genres.map(::normalize).toSet()
+        return when {
+            preferredGenres.isEmpty() -> BASE_DIVERSITY_SCORE
+            filmGenres.none { it in preferredGenres } -> HIGH_DIVERSITY_SCORE
+            filmGenres.size > 1 -> MEDIUM_DIVERSITY_SCORE
+            else -> LOW_DIVERSITY_SCORE
+        }
+    }
+
+    private fun inferredMoods(film: Film): Set<String> {
+        val text = normalize("${film.title} ${film.description} ${film.genres.joinToString(" ")}")
+        return moodLexicon
+            .filterValues { keywords -> keywords.any { keyword -> text.contains(keyword) } }
+            .keys
+    }
+
+    private fun matchingGenres(
+        film: Film,
+        preferences: UserPreferences?,
+    ): List<String> {
+        val filmGenres = film.genres.associateBy { normalize(it) }
+        return preferences
+            ?.weightedGenres
+            .orEmpty()
+            .keys
+            .map(::normalize)
+            .mapNotNull { filmGenres[it] }
+    }
+
+    private fun matchingPeople(
+        film: Film,
+        preferences: UserPreferences?,
+    ): List<String> {
+        val people = (film.cast + film.directors).associateBy { normalize(it) }
+        return preferences
+            ?.castAndDirectors
+            .orEmpty()
+            .map(::normalize)
+            .mapNotNull { people[it] }
+    }
+
+    private fun distribute(
+        vector: MutableSparseVector,
+        namespace: String,
+        values: Collection<String>,
+        totalWeight: Double,
+    ) {
+        val uniqueValues = values.map(::normalize).filter { it.isNotBlank() }.distinct()
+        if (uniqueValues.isEmpty()) {
+            return
+        }
+        val itemWeight = totalWeight / uniqueValues.size
+        uniqueValues.forEach { vector.add(feature(namespace, it), itemWeight) }
+    }
+
+    private fun ratingSignal(score: Int): Double =
+        when (score.coerceIn(MIN_USER_RATING, MAX_USER_RATING)) {
+            10 -> 1.0
+            9 -> 0.9
+            8 -> 0.7
+            7 -> 0.4
+            6 -> 0.1
+            5 -> 0.0
+            4 -> -0.3
+            3 -> -0.5
+            2 -> -0.8
+            else -> -1.0
+        }
+
+    private fun normalizeRating(rating: Double): Double = (rating / MAX_RATING_VALUE).coerceIn(0.0, 1.0)
+
+    private fun decadeOf(year: Int): String = "${year / 10 * 10}s"
+
+    private fun tokenize(text: String): List<String> =
+        normalize(text)
+            .split(tokenSeparatorRegex)
+            .asSequence()
+            .filter { it.length >= MIN_TOKEN_LENGTH }
+            .filterNot { it in stopWords }
+            .distinct()
+            .toList()
+
+    private fun feature(
+        namespace: String,
+        value: String,
+    ): String = "$namespace:${normalize(value)}"
+
+    private fun normalize(value: String): String =
+        value
+            .trim()
+            .lowercase(Locale.getDefault())
+
+    private fun cosineSimilarity(
+        left: SparseVector,
+        right: SparseVector,
+    ): Double {
+        if (left.values.isEmpty() || right.values.isEmpty()) {
+            return 0.0
+        }
+
+        val dot =
+            left.values
+                .entries
+                .sumOf { (feature, weight) -> weight * (right.values[feature] ?: 0.0) }
+        val leftNorm = sqrt(left.values.values.sumOf { it * it })
+        val rightNorm = sqrt(right.values.values.sumOf { it * it })
+        if (leftNorm == 0.0 || rightNorm == 0.0) {
+            return 0.0
+        }
+
+        return dot / (leftNorm * rightNorm)
+    }
+
+    private fun roundScore(score: Double): Double =
+        kotlin.math.round(score * SCORE_ROUNDING_FACTOR) / SCORE_ROUNDING_FACTOR
+
+    private fun Iterable<Double>.averageOrNull(): Double? {
+        val values = toList()
+        return values.takeIf { it.isNotEmpty() }?.average()
+    }
+
+    private data class ScoredRecommendation(
+        val result: RecommendationResult,
+        val relevanceScore: Double,
+        val qualityScore: Double,
+        val contextScore: Double,
+        val noveltyScore: Double,
+        val diversityScore: Double,
+    )
+
+    private data class ScoreContributions(
+        val relevance: Double,
+        val quality: Double,
+        val context: Double,
+        val novelty: Double,
+        val diversity: Double,
+    )
+
+    private data class SparseVector(
+        val values: Map<String, Double>,
+    ) {
+        fun scale(weight: Double): SparseVector = SparseVector(values.mapValues { it.value * weight })
+    }
+
+    private class MutableSparseVector {
+        private val values = mutableMapOf<String, Double>()
+
+        fun add(
+            feature: String,
+            weight: Double,
+        ) {
+            if (weight == 0.0) {
+                return
+            }
+            values[feature] = (values[feature] ?: 0.0) + weight
+        }
+
+        fun add(vector: SparseVector) {
+            vector.values.forEach { (feature, weight) -> add(feature, weight) }
+        }
+
+        fun toSparseVector(): SparseVector = SparseVector(values.filterValues { it != 0.0 })
+    }
+
+    private companion object {
+        private const val RECOMMENDATION_COMPLETED_LOG =
+            "Recommendation request completed: userId='{}', contentType='{}', moodPresent={}, " +
+                "libraryOnly={}, limit={}, candidatesCount={}, returnedCount={}"
+        private const val RECOMMENDATION_FEEDBACK_SAVED_LOG =
+            "Recommendation feedback saved: userId='{}', filmId='{}', eventType='{}'"
+        private const val RECOMMENDATION_WEIGHTS_UPDATED_LOG =
+            "Recommendation weights updated: userId='{}', eventType='{}', oldWeightsHash={}, newWeightsHash={}"
+
+        private const val MAX_PREFERENCE_WEIGHT = 5.0
+        private const val MAX_RATING_VALUE = 10.0
+        private const val MIN_USER_RATING = 1
+        private const val MAX_USER_RATING = 10
+        private const val MIN_TOKEN_LENGTH = 3
+        private const val MAX_REASON_ITEMS = 2
+        private const val SCORE_ROUNDING_FACTOR = 1000.0
+
+        private const val PREFERENCE_PLOT_WEIGHT = 0.6
+        private const val PREFERENCE_ERA_WEIGHT = 0.7
+        private const val PREFERENCE_PERSON_WEIGHT = 0.8
+        private const val PREFERENCE_MOOD_WEIGHT = 0.8
+        private const val PREFERENCE_CONTENT_TYPE_WEIGHT = 0.5
+        private const val LIBRARY_SIGNAL_WEIGHT = 0.25
+
+        private const val LEARNING_RATE = 0.03
+
+        private const val LIBRARY_NOVELTY_SCORE = 0.85
+        private const val CATALOG_NOVELTY_SCORE = 0.65
+        private const val BASE_CONTEXT_SCORE = 0.5
+        private const val BASE_QUALITY_SCORE = 0.5
+        private const val BASE_DIVERSITY_SCORE = 0.5
+        private const val HIGH_DIVERSITY_SCORE = 1.0
+        private const val MEDIUM_DIVERSITY_SCORE = 0.6
+        private const val LOW_DIVERSITY_SCORE = 0.3
+        private const val STRONG_REASON_THRESHOLD = 0.15
+        private const val QUALITY_REASON_THRESHOLD = 0.75
+
+        private val tokenSeparatorRegex = Regex("[^\\p{L}0-9]+")
+        private val stopWords =
+            setOf(
+                "and",
+                "the",
+                "for",
+                "with",
+                "about",
+                "into",
+                "from",
+            )
+        private val moodLexicon =
+            mapOf(
+                "tense" to listOf("thriller", "suspense", "tension", "rescue", "crime"),
+                "slow-burn" to listOf("slow", "meditative", "grounded"),
+                "feel-good" to listOf("comedy", "family", "summer", "kind", "warm"),
+                "dark" to listOf("dark", "noir", "horror", "murder", "crime"),
+                "romantic" to listOf("romance", "love", "relationship"),
+                "focused" to listOf("science", "mission", "detective", "investigation", "sci-fi"),
+            )
     }
 }

src/main/kotlin/com/project/movienight/application/services/UserRecommendationWeightsService.kt

@@ -0,0 +1,51 @@
+package com.project.movienight.application.services
+
+import com.project.movienight.application.ports.input.GetUserRecommendationWeightsUseCase
+import com.project.movienight.application.ports.input.UpdateUserRecommendationWeightsCommand
+import com.project.movienight.application.ports.input.UpdateUserRecommendationWeightsUseCase
+import com.project.movienight.application.ports.output.UserRecommendationWeightsRepositoryPort
+import com.project.movienight.application.ports.output.UserRepositoryPort
+import com.project.movienight.domain.exception.EntityNotFoundException
+import com.project.movienight.domain.model.UserRecommendationWeights
+import org.springframework.stereotype.Service
+import java.util.UUID
+
+@Service
+class UserRecommendationWeightsService(
+    private val userRecommendationWeightsRepository: UserRecommendationWeightsRepositoryPort,
+    private val userRepository: UserRepositoryPort,
+) : GetUserRecommendationWeightsUseCase,
+    UpdateUserRecommendationWeightsUseCase {
+    override fun get(userId: UUID): UserRecommendationWeights {
+        ensureUserExists(userId)
+        return (
+            userRecommendationWeightsRepository.findByUserId(userId)
+                ?: UserRecommendationWeights.defaultFor(userId)
+        ).normalized()
+    }
+
+    override fun update(command: UpdateUserRecommendationWeightsCommand): UserRecommendationWeights {
+        ensureUserExists(command.userId)
+        return userRecommendationWeightsRepository.save(
+            UserRecommendationWeights(
+                userId = command.userId,
+                relevanceWeight = command.relevanceWeight,
+                qualityWeight = command.qualityWeight,
+                contextWeight = command.contextWeight,
+                noveltyWeight = command.noveltyWeight,
+                diversityWeight = command.diversityWeight,
+                genreVectorWeight = command.genreVectorWeight,
+                plotVectorWeight = command.plotVectorWeight,
+                moodVectorWeight = command.moodVectorWeight,
+                eraVectorWeight = command.eraVectorWeight,
+                peopleVectorWeight = command.peopleVectorWeight,
+                contentTypeVectorWeight = command.contentTypeVectorWeight,
+            ),
+        )
+    }
+
+    private fun ensureUserExists(userId: UUID) {
+        userRepository.findById(userId)
+            ?: throw EntityNotFoundException(entity = "User", id = userId.toString())
+    }
+}

src/main/kotlin/com/project/movienight/config/JellyfinIntegrationProperties.kt

@@ -6,6 +6,7 @@ import org.springframework.boot.context.properties.ConfigurationProperties
 data class JellyfinIntegrationProperties(
     val enabled: Boolean = false,
     val baseUrl: String = "",
+    val webUrl: String = "",
     val apiKey: String = "",
     val syncIntervalMs: Long = 1_800_000,
     val requestTimeoutMs: Long = 20_000,

src/main/kotlin/com/project/movienight/domain/model/RecommendationContext.kt

@@ -6,6 +6,7 @@ data class RecommendationContext(
     val userId: UUID,
     val contentType: ContentType? = null,
     val mood: String? = null,
+    val libraryOnly: Boolean = false,
     val limit: Int = 10,
 )
 

src/main/kotlin/com/project/movienight/domain/model/RecommendationEvent.kt

@@ -0,0 +1,24 @@
+package com.project.movienight.domain.model
+
+import java.time.LocalDateTime
+import java.util.UUID
+
+data class RecommendationEvent(
+    val id: UUID,
+    val userId: UUID,
+    val filmId: UUID,
+    val eventType: RecommendationEventType,
+    val score: Double? = null,
+    val relevanceScore: Double? = null,
+    val qualityScore: Double? = null,
+    val contextScore: Double? = null,
+    val noveltyScore: Double? = null,
+    val diversityScore: Double? = null,
+    val createdAt: LocalDateTime = LocalDateTime.now(),
+)
+
+enum class RecommendationEventType {
+    RECOMMENDED,
+    ACCEPTED,
+    REJECTED,
+}

src/main/kotlin/com/project/movienight/domain/model/RecommendationStyle.kt

@@ -0,0 +1,9 @@
+package com.project.movienight.domain.model
+
+enum class RecommendationStyle {
+    BALANCED,
+    QUALITY_FIRST,
+    MOOD_FIRST,
+    DISCOVERY,
+    SIMILAR_TO_FAVORITES,
+}

src/main/kotlin/com/project/movienight/domain/model/UserRecommendationWeights.kt

@@ -0,0 +1,233 @@
+package com.project.movienight.domain.model
+
+import java.time.LocalDateTime
+import java.util.UUID
+
+data class UserRecommendationWeights(
+    val userId: UUID,
+    val relevanceWeight: Double = DEFAULT_RELEVANCE_WEIGHT,
+    val qualityWeight: Double = DEFAULT_QUALITY_WEIGHT,
+    val contextWeight: Double = DEFAULT_CONTEXT_WEIGHT,
+    val noveltyWeight: Double = DEFAULT_NOVELTY_WEIGHT,
+    val diversityWeight: Double = DEFAULT_DIVERSITY_WEIGHT,
+    val genreVectorWeight: Double = DEFAULT_GENRE_VECTOR_WEIGHT,
+    val plotVectorWeight: Double = DEFAULT_PLOT_VECTOR_WEIGHT,
+    val moodVectorWeight: Double = DEFAULT_MOOD_VECTOR_WEIGHT,
+    val eraVectorWeight: Double = DEFAULT_ERA_VECTOR_WEIGHT,
+    val peopleVectorWeight: Double = DEFAULT_PEOPLE_VECTOR_WEIGHT,
+    val contentTypeVectorWeight: Double = DEFAULT_CONTENT_TYPE_VECTOR_WEIGHT,
+    val updatedAt: LocalDateTime = LocalDateTime.now(),
+) {
+    fun normalized(updatedAt: LocalDateTime = this.updatedAt): UserRecommendationWeights {
+        val scoreWeights =
+            normalizeBounded(
+                values =
+                    listOf(
+                        relevanceWeight,
+                        qualityWeight,
+                        contextWeight,
+                        noveltyWeight,
+                        diversityWeight,
+                    ),
+                defaults = DEFAULT_SCORE_WEIGHTS,
+                min = MIN_SCORE_WEIGHT,
+                max = MAX_SCORE_WEIGHT,
+            )
+        val vectorWeights =
+            normalizeBounded(
+                values =
+                    listOf(
+                        genreVectorWeight,
+                        plotVectorWeight,
+                        moodVectorWeight,
+                        eraVectorWeight,
+                        peopleVectorWeight,
+                        contentTypeVectorWeight,
+                    ),
+                defaults = DEFAULT_VECTOR_WEIGHTS,
+                min = MIN_VECTOR_WEIGHT,
+                max = MAX_VECTOR_WEIGHT,
+            )
+
+        return copy(
+            relevanceWeight = scoreWeights[0],
+            qualityWeight = scoreWeights[1],
+            contextWeight = scoreWeights[2],
+            noveltyWeight = scoreWeights[3],
+            diversityWeight = scoreWeights[4],
+            genreVectorWeight = vectorWeights[0],
+            plotVectorWeight = vectorWeights[1],
+            moodVectorWeight = vectorWeights[2],
+            eraVectorWeight = vectorWeights[3],
+            peopleVectorWeight = vectorWeights[4],
+            contentTypeVectorWeight = vectorWeights[5],
+            updatedAt = updatedAt,
+        )
+    }
+
+    companion object {
+        const val DEFAULT_RELEVANCE_WEIGHT = 0.55
+        const val DEFAULT_QUALITY_WEIGHT = 0.15
+        const val DEFAULT_CONTEXT_WEIGHT = 0.10
+        const val DEFAULT_NOVELTY_WEIGHT = 0.10
+        const val DEFAULT_DIVERSITY_WEIGHT = 0.10
+
+        const val DEFAULT_GENRE_VECTOR_WEIGHT = 0.25
+        const val DEFAULT_PLOT_VECTOR_WEIGHT = 0.35
+        const val DEFAULT_MOOD_VECTOR_WEIGHT = 0.15
+        const val DEFAULT_ERA_VECTOR_WEIGHT = 0.10
+        const val DEFAULT_PEOPLE_VECTOR_WEIGHT = 0.10
+        const val DEFAULT_CONTENT_TYPE_VECTOR_WEIGHT = 0.05
+
+        const val MIN_SCORE_WEIGHT = 0.05
+        const val MAX_SCORE_WEIGHT = 0.75
+        const val MIN_VECTOR_WEIGHT = 0.03
+        const val MAX_VECTOR_WEIGHT = 0.60
+
+        private val DEFAULT_SCORE_WEIGHTS =
+            listOf(
+                DEFAULT_RELEVANCE_WEIGHT,
+                DEFAULT_QUALITY_WEIGHT,
+                DEFAULT_CONTEXT_WEIGHT,
+                DEFAULT_NOVELTY_WEIGHT,
+                DEFAULT_DIVERSITY_WEIGHT,
+            )
+        private val DEFAULT_VECTOR_WEIGHTS =
+            listOf(
+                DEFAULT_GENRE_VECTOR_WEIGHT,
+                DEFAULT_PLOT_VECTOR_WEIGHT,
+                DEFAULT_MOOD_VECTOR_WEIGHT,
+                DEFAULT_ERA_VECTOR_WEIGHT,
+                DEFAULT_PEOPLE_VECTOR_WEIGHT,
+                DEFAULT_CONTENT_TYPE_VECTOR_WEIGHT,
+            )
+
+        fun defaultFor(userId: UUID): UserRecommendationWeights = UserRecommendationWeights(userId = userId)
+
+        fun forStyle(
+            userId: UUID,
+            style: RecommendationStyle,
+        ): UserRecommendationWeights =
+            when (style) {
+                RecommendationStyle.BALANCED -> {
+                    defaultFor(userId)
+                }
+
+                RecommendationStyle.QUALITY_FIRST -> {
+                    UserRecommendationWeights(
+                        userId = userId,
+                        relevanceWeight = 0.40,
+                        qualityWeight = 0.35,
+                        contextWeight = 0.10,
+                        noveltyWeight = 0.05,
+                        diversityWeight = 0.10,
+                    )
+                }
+
+                RecommendationStyle.MOOD_FIRST -> {
+                    UserRecommendationWeights(
+                        userId = userId,
+                        relevanceWeight = 0.45,
+                        qualityWeight = 0.10,
+                        contextWeight = 0.25,
+                        noveltyWeight = 0.10,
+                        diversityWeight = 0.10,
+                        moodVectorWeight = 0.30,
+                    )
+                }
+
+                RecommendationStyle.DISCOVERY -> {
+                    UserRecommendationWeights(
+                        userId = userId,
+                        relevanceWeight = 0.30,
+                        qualityWeight = 0.10,
+                        contextWeight = 0.10,
+                        noveltyWeight = 0.25,
+                        diversityWeight = 0.25,
+                    )
+                }
+
+                RecommendationStyle.SIMILAR_TO_FAVORITES -> {
+                    UserRecommendationWeights(
+                        userId = userId,
+                        relevanceWeight = 0.70,
+                        qualityWeight = 0.10,
+                        contextWeight = 0.10,
+                        noveltyWeight = 0.05,
+                        diversityWeight = 0.05,
+                        genreVectorWeight = 0.30,
+                        plotVectorWeight = 0.40,
+                        peopleVectorWeight = 0.15,
+                    )
+                }
+            }.normalized()
+
+        private fun normalizeBounded(
+            values: List<Double>,
+            defaults: List<Double>,
+            min: Double,
+            max: Double,
+        ): List<Double> {
+            val sanitized = values.map { value -> if (value.isFinite() && value > 0.0) value else 0.0 }
+            val source = sanitized.takeIf { it.sum() > 0.0 } ?: defaults
+            val normalized = source.map { it / source.sum() }
+            return projectToBounds(normalized, min, max)
+        }
+
+        private fun projectToBounds(
+            values: List<Double>,
+            min: Double,
+            max: Double,
+        ): List<Double> {
+            val result = values.map { it.coerceIn(min, max) }.toMutableList()
+            var iterations = 0
+            var adjusting = true
+
+            while (iterations < values.size * 2 && adjusting) {
+                iterations += 1
+                val diff = 1.0 - result.sum()
+                if (kotlin.math.abs(diff) <= NORMALIZATION_EPSILON) {
+                    adjusting = false
+                } else {
+                    adjusting = redistribute(result, diff, min, max)
+                }
+            }
+
+            return result
+        }
+
+        private fun redistribute(
+            result: MutableList<Double>,
+            diff: Double,
+            min: Double,
+            max: Double,
+        ): Boolean =
+            if (diff > 0.0) {
+                val candidates = result.indices.filter { result[it] < max }
+                val capacity = candidates.sumOf { max - result[it] }
+                if (capacity > 0.0) {
+                    candidates.forEach { index ->
+                        val increment = diff * ((max - result[index]) / capacity)
+                        result[index] = (result[index] + increment).coerceAtMost(max)
+                    }
+                    true
+                } else {
+                    false
+                }
+            } else {
+                val candidates = result.indices.filter { result[it] > min }
+                val capacity = candidates.sumOf { result[it] - min }
+                if (capacity > 0.0) {
+                    candidates.forEach { index ->
+                        val decrement = -diff * ((result[index] - min) / capacity)
+                        result[index] = (result[index] - decrement).coerceAtLeast(min)
+                    }
+                    true
+                } else {
+                    false
+                }
+            }
+
+        private const val NORMALIZATION_EPSILON = 0.0000001
+    }
+}

src/main/resources/application.yaml

@@ -99,6 +99,7 @@ integrations:
   jellyfin:
     enabled: ${JELLYFIN_SYNC_ENABLED:false}
     base-url: ${JELLYFIN_BASE_URL:}
+    web-url: ${JELLYFIN_WEB_URL:${JELLYFIN_BASE_URL:}}
     api-key: ${JELLYFIN_API_KEY:}
     sync-interval-ms: ${JELLYFIN_SYNC_INTERVAL_MS:1800000}
     request-timeout-ms: ${JELLYFIN_REQUEST_TIMEOUT_MS:20000}

src/main/resources/db/migration/V7__recommendation_events.sql

@@ -0,0 +1,19 @@
+CREATE TABLE IF NOT EXISTS public.recommendation_events (
+    id UUID PRIMARY KEY,
+    user_id UUID NOT NULL,
+    film_id UUID NOT NULL,
+    event_type VARCHAR(64) NOT NULL,
+    score DOUBLE PRECISION,
+    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    CONSTRAINT recommendation_events_user_fk FOREIGN KEY (user_id) REFERENCES public.users(id) ON DELETE CASCADE,
+    CONSTRAINT recommendation_events_film_fk FOREIGN KEY (film_id) REFERENCES public.films(id) ON DELETE CASCADE
+);
+
+CREATE INDEX IF NOT EXISTS idx_recommendation_events_user_created
+    ON public.recommendation_events(user_id, created_at DESC);
+
+CREATE INDEX IF NOT EXISTS idx_recommendation_events_film
+    ON public.recommendation_events(film_id);
+
+CREATE INDEX IF NOT EXISTS idx_recommendation_events_type
+    ON public.recommendation_events(event_type);

src/main/resources/db/migration/V8__user_recommendation_weights.sql

@@ -0,0 +1,35 @@
+CREATE TABLE IF NOT EXISTS public.user_recommendation_weights (
+    user_id UUID PRIMARY KEY,
+    relevance_weight DOUBLE PRECISION NOT NULL DEFAULT 0.55,
+    quality_weight DOUBLE PRECISION NOT NULL DEFAULT 0.15,
+    context_weight DOUBLE PRECISION NOT NULL DEFAULT 0.10,
+    novelty_weight DOUBLE PRECISION NOT NULL DEFAULT 0.10,
+    diversity_weight DOUBLE PRECISION NOT NULL DEFAULT 0.10,
+    genre_vector_weight DOUBLE PRECISION NOT NULL DEFAULT 0.25,
+    plot_vector_weight DOUBLE PRECISION NOT NULL DEFAULT 0.35,
+    mood_vector_weight DOUBLE PRECISION NOT NULL DEFAULT 0.15,
+    era_vector_weight DOUBLE PRECISION NOT NULL DEFAULT 0.10,
+    people_vector_weight DOUBLE PRECISION NOT NULL DEFAULT 0.10,
+    content_type_vector_weight DOUBLE PRECISION NOT NULL DEFAULT 0.05,
+    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    CONSTRAINT user_recommendation_weights_user_fk
+        FOREIGN KEY (user_id) REFERENCES public.users(id) ON DELETE CASCADE
+);
+
+ALTER TABLE public.recommendation_events
+    ADD COLUMN IF NOT EXISTS relevance_score DOUBLE PRECISION;
+
+ALTER TABLE public.recommendation_events
+    ADD COLUMN IF NOT EXISTS quality_score DOUBLE PRECISION;
+
+ALTER TABLE public.recommendation_events
+    ADD COLUMN IF NOT EXISTS context_score DOUBLE PRECISION;
+
+ALTER TABLE public.recommendation_events
+    ADD COLUMN IF NOT EXISTS novelty_score DOUBLE PRECISION;
+
+ALTER TABLE public.recommendation_events
+    ADD COLUMN IF NOT EXISTS diversity_score DOUBLE PRECISION;
+
+CREATE INDEX IF NOT EXISTS idx_recommendation_events_user_film_type_created
+    ON public.recommendation_events(user_id, film_id, event_type, created_at DESC);

src/test/kotlin/com/project/movienight/RecommendationSmokeTest.kt

@@ -4,8 +4,12 @@ import com.fasterxml.jackson.databind.ObjectMapper
 import com.project.movienight.adapters.web.dto.request.CreateFilmRequest
 import com.project.movienight.adapters.web.dto.request.CreateUserRequest
 import com.project.movienight.adapters.web.dto.request.RateFilmRequest
+import com.project.movienight.adapters.web.dto.request.RecommendationOnboardingRequest
+import com.project.movienight.adapters.web.dto.request.UpdateUserRecommendationWeightsRequest
 import com.project.movienight.adapters.web.dto.request.UpsertUserPreferencesRequest
 import org.junit.jupiter.api.AfterEach
+import org.junit.jupiter.api.Assertions.assertNotEquals
+import org.junit.jupiter.api.Assertions.assertTrue
 import org.junit.jupiter.api.BeforeEach
 import org.junit.jupiter.api.Test
 import org.springframework.beans.factory.annotation.Autowired
@@ -76,6 +80,7 @@ class RecommendationSmokeTest {
                             imdbRating = 8.7,
                             platformRating = 9.0,
                             externalUrl = "https://example.com/orbital-drift",
+                            jellyfinItemId = "orbital-drift-item",
                         ),
                     )
             }.andExpect {
@@ -109,6 +114,39 @@ class RecommendationSmokeTest {
         val firstFilmId = filmIdByTitle.getValue("Orbital Drift")
         val secondFilmId = filmIdByTitle.getValue("Small Town Summer")
 
+        mockMvc
+            .get("/api/users/$userId/recommendation-weights")
+            .andExpect {
+                status { isOk() }
+                jsonPath("$.relevanceWeight") { value(0.55) }
+                jsonPath("$.plotVectorWeight") { value(0.35) }
+            }
+
+        mockMvc
+            .put("/api/users/$userId/recommendation-weights") {
+                contentType = MediaType.APPLICATION_JSON
+                content =
+                    objectMapper.writeValueAsString(
+                        UpdateUserRecommendationWeightsRequest(
+                            relevanceWeight = 0.60,
+                            qualityWeight = 0.10,
+                            contextWeight = 0.15,
+                            noveltyWeight = 0.10,
+                            diversityWeight = 0.05,
+                            genreVectorWeight = 0.30,
+                            plotVectorWeight = 0.30,
+                            moodVectorWeight = 0.20,
+                            eraVectorWeight = 0.05,
+                            peopleVectorWeight = 0.10,
+                            contentTypeVectorWeight = 0.05,
+                        ),
+                    )
+            }.andExpect {
+                status { isOk() }
+                jsonPath("$.relevanceWeight") { value(0.6) }
+                jsonPath("$.genreVectorWeight") { value(0.3) }
+            }
+
         mockMvc
             .put("/api/users/$userId/preferences") {
                 contentType = MediaType.APPLICATION_JSON
@@ -154,15 +192,218 @@ class RecommendationSmokeTest {
                 param("limit", "2")
             }.andExpect {
                 status { isOk() }
+                jsonPath("$[0].filmId") { value(firstFilmId.toString()) }
                 jsonPath("$[0].film.id") { value(firstFilmId.toString()) }
+                jsonPath("$[0].watchUrl") {
+                    value("https://jellyfin.example.test/web/#/details?id=orbital-drift-item")
+                }
+                jsonPath("$[0].reasons[0]") { exists() }
+            }
+
+        val recommendedBreakdownCount =
+            jdbcTemplate.queryForObject(
+                """
+                SELECT COUNT(*)
+                FROM recommendation_events
+                WHERE user_id = ?
+                  AND film_id = ?
+                  AND event_type = 'RECOMMENDED'
+                  AND relevance_score IS NOT NULL
+                  AND quality_score IS NOT NULL
+                """.trimIndent(),
+                Int::class.java,
+                userId,
+                firstFilmId,
+            )
+        assertTrue((recommendedBreakdownCount ?: 0) > 0)
+
+        val weightsBeforeFeedback = findScoreWeights(userId)
+
+        mockMvc
+            .post("/api/users/$userId/recommendations/$firstFilmId/accept")
+            .andExpect {
+                status { isOk() }
+                jsonPath("$.filmId") { value(firstFilmId.toString()) }
+                jsonPath("$.eventType") { value("ACCEPTED") }
+                jsonPath("$.relevanceScore") { exists() }
+            }
+
+        val weightsAfterAccept = findScoreWeights(userId)
+        assertNotEquals(weightsBeforeFeedback, weightsAfterAccept)
+        assertTrue(weightsAfterAccept.all { it in 0.05..0.75 })
+
+        mockMvc
+            .post("/api/users/$userId/recommendations/$firstFilmId/reject")
+            .andExpect {
+                status { isOk() }
+                jsonPath("$.filmId") { value(firstFilmId.toString()) }
+                jsonPath("$.eventType") { value("REJECTED") }
+            }
+
+        mockMvc
+            .get("/api/users/$userId/recommendations") {
+                param("contentType", "FILM")
+                param("libraryOnly", "true")
+                param("limit", "2")
+            }.andExpect {
+                status { isOk() }
+                jsonPath("$") { isEmpty() }
+            }
+    }
+
+    @Test
+    fun `should complete recommendation onboarding`() {
+        mockMvc
+            .post("/api/users") {
+                contentType = MediaType.APPLICATION_JSON
+                content = objectMapper.writeValueAsString(CreateUserRequest(name = "Alex", email = "alex@example.com"))
+            }.andExpect {
+                status { isCreated() }
+            }
+
+        val userId =
+            UUID.fromString(
+                jdbcTemplate.queryForObject(
+                    "SELECT id FROM users WHERE email = ?",
+                    String::class.java,
+                    "alex@example.com",
+                ),
+            )
+
+        val likedFilmId = createFilm(title = "Neon Rescue", genres = listOf("SCI-FI"), imdbRating = 8.8)
+        val dislikedFilmId = createFilm(title = "Quiet Village", genres = listOf("DRAMA"), imdbRating = 5.0)
+        val libraryFilmId = createFilm(title = "Space Trial", genres = listOf("SCI-FI"), imdbRating = 7.8)
+        val watchedFilmId = createFilm(title = "Old Mission", genres = listOf("THRILLER"), imdbRating = 8.1)
+
+        mockMvc
+            .post("/api/users/$userId/recommendation-onboarding") {
+                contentType = MediaType.APPLICATION_JSON
+                content =
+                    objectMapper.writeValueAsString(
+                        RecommendationOnboardingRequest(
+                            weightedGenres = mapOf("SCI-FI" to 5, "THRILLER" to 3),
+                            moods = listOf("focused", "tense"),
+                            contentTypes = listOf("FILM"),
+                            likedFilmIds = listOf(likedFilmId),
+                            dislikedFilmIds = listOf(dislikedFilmId),
+                            libraryFilmIds = listOf(libraryFilmId),
+                            watchedFilmIds = listOf(watchedFilmId),
+                            recommendationStyle = "DISCOVERY",
+                        ),
+                    )
+            }.andExpect {
+                status { isOk() }
+                jsonPath("$.preferences.weightedGenres['SCI-FI']") { value(5) }
+                jsonPath("$.weights.noveltyWeight") { value(0.25) }
+                jsonPath("$.weights.diversityWeight") { value(0.25) }
+                jsonPath("$.likedFilmsCount") { value(1) }
+                jsonPath("$.dislikedFilmsCount") { value(1) }
+                jsonPath("$.libraryFilmsCount") { value(1) }
+                jsonPath("$.watchedFilmsCount") { value(1) }
+            }
+
+        assertDatabaseCount(
+            """
+            SELECT COUNT(*)
+            FROM film_ratings
+            WHERE user_id = ?
+              AND film_id IN (?, ?)
+            """.trimIndent(),
+            userId,
+            likedFilmId,
+            dislikedFilmId,
+        )
+        assertDatabaseCount(
+            """
+            SELECT COUNT(*)
+            FROM favorites
+            WHERE user_id = ?
+              AND film_id = ?
+              AND is_viewed = TRUE
+            """.trimIndent(),
+            userId,
+            watchedFilmId,
+        )
+
+        mockMvc
+            .get("/api/users/$userId/recommendations") {
+                param("contentType", "FILM")
+                param("limit", "3")
+            }.andExpect {
+                status { isOk() }
+                jsonPath("$[0].reasons[0]") { exists() }
             }
     }
 
     private fun cleanDatabase() {
+        jdbcTemplate.execute("DELETE FROM recommendation_events")
+        jdbcTemplate.execute("DELETE FROM user_recommendation_weights")
         jdbcTemplate.execute("DELETE FROM film_ratings")
         jdbcTemplate.execute("DELETE FROM user_preferences")
         jdbcTemplate.execute("DELETE FROM favorites")
         jdbcTemplate.execute("DELETE FROM films")
         jdbcTemplate.execute("DELETE FROM users")
     }
+
+    private fun findScoreWeights(userId: UUID): List<Double> =
+        jdbcTemplate
+            .queryForMap(
+                """
+                SELECT relevance_weight,
+                       quality_weight,
+                       context_weight,
+                       novelty_weight,
+                       diversity_weight
+                FROM user_recommendation_weights
+                WHERE user_id = ?
+                """.trimIndent(),
+                userId,
+            ).let { row ->
+                listOf(
+                    row.getValue("RELEVANCE_WEIGHT"),
+                    row.getValue("QUALITY_WEIGHT"),
+                    row.getValue("CONTEXT_WEIGHT"),
+                    row.getValue("NOVELTY_WEIGHT"),
+                    row.getValue("DIVERSITY_WEIGHT"),
+                ).map { (it as Number).toDouble() }
+            }
+
+    private fun createFilm(
+        title: String,
+        genres: List<String>,
+        imdbRating: Double,
+    ): UUID {
+        mockMvc
+            .post("/api/films") {
+                contentType = MediaType.APPLICATION_JSON
+                content =
+                    objectMapper.writeValueAsString(
+                        CreateFilmRequest(
+                            title = title,
+                            description = "$title description",
+                            contentType = "FILM",
+                            genres = genres,
+                            imdbRating = imdbRating,
+                        ),
+                    )
+            }.andExpect {
+                status { isCreated() }
+            }
+
+        return UUID.fromString(
+            jdbcTemplate.queryForObject(
+                "SELECT id FROM films WHERE title = ?",
+                String::class.java,
+                title,
+            ),
+        )
+    }
+
+    private fun assertDatabaseCount(
+        sql: String,
+        vararg args: Any,
+    ) {
+        val count = jdbcTemplate.queryForObject(sql, Int::class.java, *args)
+        assertTrue((count ?: 0) > 0)
+    }
 }

src/test/kotlin/com/project/movienight/domain/model/UserRecommendationWeightsTest.kt

@@ -0,0 +1,74 @@
+package com.project.movienight.domain.model
+
+import org.junit.jupiter.api.Assertions.assertEquals
+import org.junit.jupiter.api.Assertions.assertTrue
+import org.junit.jupiter.api.Test
+import java.util.UUID
+
+class UserRecommendationWeightsTest {
+    @Test
+    fun `should keep default weights normalized`() {
+        val weights = UserRecommendationWeights.defaultFor(UUID.randomUUID()).normalized()
+
+        assertEquals(1.0, weights.scoreWeightSum(), EPSILON)
+        assertEquals(1.0, weights.vectorWeightSum(), EPSILON)
+        assertEquals(0.55, weights.relevanceWeight, EPSILON)
+        assertEquals(0.35, weights.plotVectorWeight, EPSILON)
+    }
+
+    @Test
+    fun `should normalize and bound invalid weights`() {
+        val weights =
+            UserRecommendationWeights(
+                userId = UUID.randomUUID(),
+                relevanceWeight = 100.0,
+                qualityWeight = -5.0,
+                contextWeight = 0.0,
+                noveltyWeight = 0.0,
+                diversityWeight = 0.0,
+                genreVectorWeight = 100.0,
+                plotVectorWeight = 0.0,
+                moodVectorWeight = 0.0,
+                eraVectorWeight = 0.0,
+                peopleVectorWeight = 0.0,
+                contentTypeVectorWeight = 0.0,
+            ).normalized()
+
+        assertEquals(1.0, weights.scoreWeightSum(), EPSILON)
+        assertEquals(1.0, weights.vectorWeightSum(), EPSILON)
+        assertTrue(
+            listOf(
+                weights.relevanceWeight,
+                weights.qualityWeight,
+                weights.contextWeight,
+                weights.noveltyWeight,
+                weights.diversityWeight,
+            ).all { it in UserRecommendationWeights.MIN_SCORE_WEIGHT..UserRecommendationWeights.MAX_SCORE_WEIGHT },
+        )
+        assertTrue(
+            listOf(
+                weights.genreVectorWeight,
+                weights.plotVectorWeight,
+                weights.moodVectorWeight,
+                weights.eraVectorWeight,
+                weights.peopleVectorWeight,
+                weights.contentTypeVectorWeight,
+            ).all { it in UserRecommendationWeights.MIN_VECTOR_WEIGHT..UserRecommendationWeights.MAX_VECTOR_WEIGHT },
+        )
+    }
+
+    private fun UserRecommendationWeights.scoreWeightSum(): Double =
+        relevanceWeight + qualityWeight + contextWeight + noveltyWeight + diversityWeight
+
+    private fun UserRecommendationWeights.vectorWeightSum(): Double =
+        genreVectorWeight +
+            plotVectorWeight +
+            moodVectorWeight +
+            eraVectorWeight +
+            peopleVectorWeight +
+            contentTypeVectorWeight
+
+    private companion object {
+        private const val EPSILON = 0.000001
+    }
+}

src/test/resources/application-test.yaml

@@ -26,3 +26,7 @@ services:
       - censored
       - epstein
       - python
+
+integrations:
+  jellyfin:
+    web-url: https://jellyfin.example.test